CI Performance Gates for Flutter Physical AI Companion Apps

#flutter #ci #performance #robotics

CI Performance Gates for Flutter Physical AI Companion Apps

Performance regressions can arrive through unrelated changes: a new chart, extra JSON parsing, larger images, or more frequent state emissions. Automated gates catch some before release.

Choose a stable test setup

Use profile or release-like builds. Keep device, OS, power state, and workload consistent across comparisons.

Gate product-relevant metrics

Track frame build/raster time, startup, memory growth, decode time, queue depth, and telemetry age. Avoid one threshold for every device.

Store a baseline

Record commit, dependency lockfile, test trace, device details, and metric definitions. Compare repeated runs rather than trusting one noisy sample.

Keep tests deterministic

Replay saved telemetry and mock network timing for UI tests. Use hardware-in-the-loop tests for real camera, gateway, inference, and robot paths.

Make failures actionable

Publish slow stage, affected screen, trace ID, and before/after metrics so a failure points toward a cause.

Example implementation

This Dart snippet illustrates one part of the design. Adapt types, transport behavior, timestamp semantics, and platform APIs to your actual system before production use.

class PerformanceBudget {
  const PerformanceBudget({
    required this.maxP95FrameBuildMs,
    required this.maxP95DecodeMs,
    required this.maxP95TelemetryAgeMs,
  });
  final double maxP95FrameBuildMs;
  final double maxP95DecodeMs;
  final double maxP95TelemetryAgeMs;
}

Performance checklist

Measure p95/p99 behavior on target hardware, not only a desktop simulator.
Bound queues and buffers; define what may be coalesced or dropped.
Keep heavy parsing, image processing, and inference away from the Flutter UI thread.
Track source timestamps, sequence numbers, stale-data rate, and dropped updates.
Test network degradation, reconnects, sustained operation, and resource pressure.
Keep control commands and critical alerts separate from replaceable telemetry.